Key Takeaways
- Cornell’s IonNet AI framework predicts lithium‑ion mobility in solid electrolytes directly from chemical composition, eliminating the need for detailed crystal structures.
- IonNet screened ~4,500 stable compounds and ~5 million substituted compositions, surfacing 87 promising fast‑ion conductors and ~63,000 candidates, with 13 of 20 physics‑validated predictions confirmed as fast conductors.
- The tool serves as a fast “front‑end” that prioritizes materials before costly experiments or simulations, saving tens of thousands of CPU hours per material.
- By analyzing IonNet’s outputs, researchers extracted chemical design rules that explain why certain compositions enable rapid lithium‑ion transport.
- A complementary study on concentration batteries showed that engineering the ion‑solvent environment can raise zinc‑/copper‑based cell voltages from <0.06 V to 0.7 V—over ten times the assumed limit—enabling water‑based batteries that reach 2.2‑2.5 V when paired with conventional electrodes.
- Both projects share the goal of shifting battery design from trial‑and‑error to rational, AI‑guided design grounded in chemistry and physics.
- The work was supported by Cornell’s Eric and Wendy Schmidt AI in Science Postdoctoral Fellowship and involved collaboration with the University of Puerto Rico–Río Piedras.
Introduction and Significance of Electrolytes
Electrolytes are the ion‑conducting media that shuttle lithium ions between a battery’s electrodes, directly influencing energy density, lifetime, and charging speed. Despite their importance, designing effective electrolytes remains notoriously difficult: even minute tweaks in composition can dramatically alter ion transport, stability, and safety. “Electrolytes are not just supporting materials in batteries; they are a design space,” said Fengqi You, the Roxanne E. and Michael J. Zak Professor in Energy Systems Engineering at Cornell’s Duffield College of Engineering. This perspective drives You’s group to treat electrolyte discovery as a problem of navigating a vast chemical landscape rather than a series of ad‑hoc experiments.
Introducing IonNet
To accelerate navigation of that landscape, You’s team developed IonNet, an artificial‑intelligence framework that predicts how quickly lithium ions move through solid materials using only their chemical composition. The study describing IonNet appeared Aug. 7 in Science Advances. Unlike many material‑property models that rely on known crystal structures, IonNet operates composition‑first, making it especially useful for evaluating novel or experimentally reported compounds where structural data are scarce.
How IonNet Works Without Crystal Structures
“Most AI models for materials require reliable crystal structures, which are often unavailable for new or experimentally reported compounds,” explained You, who led the study with postdoctoral researcher and first author Zhilong Wang. IonNet sidesteps this bottleneck by learning patterns from large datasets of known ionic conductors, linking elemental identities, oxidation states, and local bonding environments to predicted ion mobility. This approach allows researchers to screen vast libraries of candidate electrolytes before investing in costly synthesis or simulation.
Results: Candidates Identified and Validation
Applying IonNet to a database of roughly 4,500 known stable compounds yielded 87 fast‑ion‑conductor prospects. When the researchers expanded the search to about five million chemically substituted compositions, the model flagged nearly 63,000 high‑potential candidates. To gauge the reliability of these predictions, the team performed physics‑based simulations on a subset of 20 materials. Thirteen of those simulations confirmed fast‑ion conduction, demonstrating IonNet’s ability to enrich the discovery pipeline with a high true‑positive rate.
AI as a Front‑End Tool
You emphasized that IonNet is not a replacement for experiment or high‑level computation but rather a rapid screening stage. “While IonNet does not replace experiments or high‑level simulations, it is a fast front‑end that prioritizes candidates before committing major experimental or computational resources,” he said. Evaluating a single material with ab‑initio molecular dynamics can consume tens of thousands of CPU hours; IonNet reduces that burden by orders of magnitude, allowing researchers to focus experimental effort on the most promising leads.
Extracting Chemical Design Rules
Beyond delivering a ranked list, IonNet was interrogated to uncover the underlying chemistry that drives high ion mobility. “We did not just want a list of promising materials,” said You, who also directs the Cornell University AI4S Initiative. “We wanted to understand the chemical principles that make lithium ions move quickly, because that is what makes AI useful for experimental design.” By analyzing feature importance and generating interpretable rules, the team identified structural motifs—such as particular polyanion frameworks and cation‑size mismatches—that consistently correlate with superior lithium transport, providing actionable guidance for chemists designing new electrolytes.
Second Study: Concentration Batteries
While Ionnet tackles solid electrolytes, a parallel project explored an unconventional energy‑storage concept: concentration batteries. These devices generate electricity from gradients in electrolyte composition rather than from differing electrode materials. Traditionally, concentration cells employing the same redox couple at both electrodes were thought to be limited to less than 0.06 volts. The Cornell team, working with researchers at the University of Puerto Rico–Río Piedras, challenged that assumption by engineering the solvation environment of zinc and copper ions within the electrolyte.
Breaking Voltage Limits in Concentration Batteries
Through precise control of ligand‑ion interactions, the researchers boosted the voltage of zinc‑/copper‑based concentration cells from the presumed ceiling to 0.7 volts—more than ten times the earlier estimate. “By engineering how ions are surrounded by molecules in the electrolyte, the researchers boosted the voltage of zinc‑ and copper‑based concentration batteries up to 0.7 volts, more than 10 times the assumed limit,” the article notes. When these high‑voltage concentration electrolytes were paired with conventional electrodes, the resulting water‑based batteries delivered operating voltages of 2.2‑2.5 volts, surpassing the typical 1.5 volts of alkaline AA batteries and opening a route to safe, aqueous energy storage with competitive performance.
Implications and Future Directions
Together, the Ionnet solid‑electrolyte work and the concentration‑battery study illustrate a unified strategy: harnessing AI, chemical insight, and physics‑based modeling to transition battery design from trial‑and‑error to rational, predictable engineering. “These electrolyte studies share the same larger goal,” You said, “to move battery design from trial and error toward rational design guided by AI, chemistry and physical insight.” The combined approach promises to accelerate the discovery of safer, higher‑energy‑density storage systems—whether solid‑state lithium batteries for electric vehicles or aqueous concentration batteries for grid‑scale applications—by drastically shortening the time between hypothesis and validation.
Funding and Collaborations
The research received partial support from Cornell’s Eric and Wendy Schmidt AI in Science Postdoctoral Fellowship, a program of Schmidt Sciences that encourages AI‑driven breakthroughs across disciplines. Collaboration extended beyond Cornell’s campus, notably with the University of Puerto Rico–Río Piedras, whose expertise in electrochemical engineering helped realize the concentration‑battery voltage enhancements. Such interdisciplinary partnerships underscore the growing consensus that advances in energy storage will rely on the seamless integration of machine learning, fundamental chemistry, and rigorous experimental validation.
https://news.cornell.edu/stories/2026/08/ai-plus-chemistry-can-expand-battery-electrolyte-design